This is my personal observation after reading lots of requests for Help.
As a new user of Fedora 42, I also experienced many of the same issues reported by other users. My Fix was using “Claude” AI.
It will assist in diagnosing your problems directly from your Browser or Desktop!
I have managed to address several issues associated with the Linux OS Some examples would be:
Nvidia driver install/fail, fixed
Davinci Resolve Studio install failed, Davinci Resolve audio corruption, fixed.
Various KDE Themes, Plasma Desktop, etc, Fixed.
Google Claude in your browser and give it a go!
Hope no one is offended by this post. also hope it helps some resolve any issues.
1 Like
Please Don’t. LLMs have absolutely no idea what they’re talking about, they are purely lying machines. If anything, I recommend searching forums (GNOME Discourse, Fedora Discourse or Reddit etc) first to see if someone have a clue on the fix. And if the search will prove fruitless, then feel free to ask the question here.
4 Likes
I am not a LLM fan by any stretch of the imagination. But I would not call them lying machines, IMO that implies bad intentions. They are simply stringing words (or tokens) together, based on mathematical probabilities.
That string of words may be useful for a given prompt and solve a problem, or it may just be utter excrement. But yeah, in here, when I see a post with “I tried … with the help of AI” or something similar, especially without clear information what the person tried at the recommendation of their chosen LLM, I usually steer clear of the question. I have no idea how much they have fouled up their system and I really don’t want to ask about every little piece of information, hoping they still remember.
3 Likes
Welcome to Fedora @bob490, as long as you report your own experience, as you made, it seams that nobody should be offended.
Please do not forget that AI takes the information from human beings made comments, which not always are correct. This is ok as long as we do have human beings which point to a solution which works and is based on trustful information.
If we start to mix the above mentioned with AI in the long run the useful information gets lost. The solutions will be based on older versions and nobody will trust the balanced information we do have now here, because being out of balance.
That is why Fedora Council is making a Proposal about the policy for AI and you can follow it below.
Just my personal observation.
4 Likes
This sounds like an ad.
3 Likes
Moved to the water cooler since this discussion seems to be more about utilizing AI to help diagnosing issues vice a question.
Please keep this civil folks.
Thanks
4 Likes
As long as the user is not linking to a specific page and give the impression with his profile, that he is affiliated with Claude, we can not do more as move it into the water cooler (off topic category). Someone did this already … You as a Community member can observe If links will appear in the future and if he creates a affiliated profile. Then it is time to start to flag and the Moderators will remove it.
1 Like
Ouch, no worries, I found assistance from using AI as a tool to assist with my issues, as well as observing the remedies it submitted! It was, and is, an opportunity for me to pick up a bit of coding, editing configs, etc. I aint a bot! Just an old man describing my fixes to Fedora. BTW, Fedora is my chosen OS as primary, having migrated from Windows!!!
3 Likes
I am happy to know that 
Sometimes you find also here in the Forum information which is tagged with an old Version of Fedora. If you do as you did, you could ask Claude if he gives you also the source of the answers. This way if it is from Fedoraproject.org you can make a own topic of each problem and link in a second response of your question to it, and mark it as solution. This way we are sure that the older information still works with the newest Versions. This would even be useful for others when they search for content here and sort it with the criteria “newest” first (as I do).
If you like, you can use this topic here to add to your profile. If someone makes researches they will see that you are aproved as a kind human being and not a 
robot
2 Likes
Welcome to the Fedora community, @bob490! 
As you can tell, some folks in the community have very strong opinions on the use of AI, but don’t take any of it personally. As long as you’re aware that AI can provide inaccurate answers to your prompts, it’s a perfectly reasonable tool to seek help from. These days I frequently use it in addition to search engines when learning new topics and have found it immensely helpful. As with all technical information from any non-official source, make sure to scrutinize it before acting on it and you’ll probably be OK.
Again, welcome to the community and don’t hesitate to ask questions here should you find yourself stuck!
6 Likes
So what’s the point of asking it in the first place? To research if it didn’t lied to you?
People on forums can also provide inaccurate or bad answers, but when someone will give you a reply like, idk, “if you want to remove French language pack, just run sudo rm -rf / --no-preserve-root in your terminal” someone will come in and rescue the situation.
It sometimes gives you answers you didn’t come up by yourself. If you then verify the answer and its sources, you might have something you would not have otherwise.
The AI should be seen in this respect like a search engine: you use it to find answers, but you should not take for granted what it does output. Everyone can write anything on the Internet, this does not make it true, nor does it prevent commands to break your system. With AI its the same: you cannot rely that an AI can calculate 1 and 1 is 2 because it is not guaranteed that by pure statistics and correlation this will be the outcome. I indeed already saw calculations in which AI did something wrong which I could have calculated by myself without technical means: that said, it actually did not calculate wrong, because it did not calculate the maths but only comparing correlations & statistics, and with its dataset and algorithms in mind, it is true that a simple mix of additions and multiplications that ended up to be 15 by maths became 19 
I also saw a lot of outputs people were using from AI that did not what they wanted to get done, and some of them check it, and if they are not 100% sure, some already opened topics here in ask.fedora, adding the AI output like a log output, and asking for help if it really does what it says (and often, it doesn’t… at the worst, in a way the users don’t experience it in an explicit way)
There are further AI uses but they are more a corporate thing, I stick with what is interesting for us here.
By the way, a machine at its lowest level can only calculate. Lies imply intention, and a machine is not capable to have that, nor to lie 
2 Likes
Btw, welcome to the community @bob490
2 Likes
Except search engines can guide you to sources you can actually trust, like Fedora Documentation or Common Issues posts on this forum. LLMs at best will give you half-done copy of that, which you would have to verify anyway. Yes, end result can possibly be the same, but then you also have to account for wasted time.
To add here, what I also saw that’s worrying me is that some people treat LLMs as a source of blessed truth, that can never be wrong.
Yes, but depending on where you ask the question, you can expect different “levels” of answers. If you go to, idk, 4chan then you can be 100% sure what you get there is either wrong or malicious. But if you ask here you can at least have some assurance that people who know something can help you/correct people who are wrong. People of course can lie or be wrong, I wouldn’t be honest if I said otherwise Here I mean that you are more likely to receive correct answers when asking fellow humans.
Personally speaking, that’s the worst thing that could happen. Maybe that’s because I’m dyslexic and English is my second language, but when I read most texts generated by LLMs I am genuinely suffering. Vast majority of the time it’s incredibly watered down text that doesn’t even say anything interesting. I already had those situations happen to me (i.e. people posting LLM output when asking for help) a couple of times and really, I feel like not posting them at all is better than posting LLM slop.
1 Like
I meant to verify the sources of an AI, which you can get. In some you can, e.g., click a button and see for each paragraph what its sources are. I indeed have seen often that the output of the AI does not correspond to its sources. In the end, it’s the same as with a search engine …
… which equals to the sources of an AI: depending on its sources, you can expect different levels of answers, and you have to verify the sources to be sure.
Here we are mostly talking of users who think about the outputs they get, and what is useful and what not. The representative topics I have seen put the AI text explicit in a box or so, or make it clear “this is the ai output:” or so, and limit it to the outputs that are relevant. Surely, much of that AI stuff could be MUCH MORE simplified, but there are worse conditions from which a problem solving discussion can rise (e.g., the “I have a problem, make something” topics). Usually, these are cases in which users also started their AI input with an expressive question, and then questioned the output. Obviously, a problem remains the users who neither ask expressive questions to the AI, at the worst containing false presumptions (that’s poison in AI queries), and then not questioning the output.
Keep in mind that there are also different AI, and their output is much influenced by the inputs. More useful outputs than the annoying “extreme” you refer to are possible.
1 Like
Hey, @tragivictoria 
To be clear, I don’t mean this in a passive aggressive way: you have strong opinions on AI and my goal wasn’t and still isn’t to change your mind. I also have my own issues with AI, but still find it to be useful in certain circumstances. So, please don’t take what I said as being personal towards you.
My goal was to make sure Bob felt welcome since this was his first post in the community and sometimes the community can feel intense or off-putting when a hot-button subject comes up.
2 Likes
I didn’t Insulted him, so I don’t get why he would feel unwelcome? He posted an advice and I said that it’s pretty bad advice and stated my reasons. Was I supposed to agree with them and move on?
Welcome to Fedora Bob!
I completely disagree with you (except of course about the fact that you found the help you need), but I’m still glad you’re here contributing your experiences
2 Likes
Interestingly I’ve just watched a talk given by Daniel Stenberg (author of curl) he gave at FrosCon back in August about the issues with AI they’ve been seeing on their project.
There’s also the other side of the same coin where Dan is stoked about some of the genuine bugs which AI has found. daniel:// stenberg://: "Joshua Rogers sent us a *massive* list of potenti…" - Mastodon
Like any tool, the quality of the finished article depends on the user of the tool being competent with its use.
1 Like